The Agent Skills Directory

[SAFE]: The skill defines a legitimate workflow for media processing, focusing on prompt construction and tool orchestration for video generation. It utilizes abstract internal tool references for animation, re-framing, and scaling tasks.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an inherent attack surface for indirect prompt injection as it ingests user-provided descriptions and brand metadata to generate prompts for video generation models.
Ingestion points: User animation requests ("anima esta foto"), uploaded media files, and local configuration files such as marca.json and marca/avatar.json.
Boundary markers: The instructions do not specify explicit delimiters or boundary markers to isolate external input from the internal prompt logic.
Capability inventory: The skill is authorized to invoke media tools (video.*, media.subir), consult costs, and write output to the trabajo/ directory.
Sanitization: Risk is mitigated through strict prompt engineering rules that require the agent to focus solely on movement descriptions (camera, subject, environment) rather than scene context, which constrains the influence of potential instructions embedded in user input.

animar